Update app.py

app.py

@@ -1,34 +1,128 @@
 import streamlit as st
-from transformers import pipeline
+st.set_page_config(page_title="Mental Health Bot", page_icon=":robot:")
+import os
+import textract
+from langchain.chat_models import ChatOpenAI
+from itertools import zip_longest
+from langchain.prompts import PromptTemplate
+from langchain.chains import LLMChain
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.vectorstores import FAISS
+from langchain.agents import initialize_agent
+from langchain.agents import AgentType
 
 
-# Load the conversational model from Hugging Face
-# chatbot = pipeline('text-generation', model='shaneperry0101/Health-Llama-3.2-1B')
-chatbot = pipeline('text-generation', model='microsoft/DialoGPT-medium')
+from langchain.agents import load_tools
+from langchain import HuggingFacePipeline
+from transformers import AutoTokenizer
+import transformers
+import torch
+os.environ['SERPAPI_API_KEY'] = '1a6691d51c19fbc52c296b7ede3b2098efff46a3e7c3bb54cb17912b0407c6ec'
 
-# Initialize session state for storing the conversation
-if 'conversation' not in st.session_state:
-    st.session_state.conversation = []
+if "llm" not in st.session_state:
+    # st.session_state["lmm"] = []
+    model = "shaneperry0101/Health-Llama-3.2-1B"
 
-# Streamlit app layout
-st.title("Chatbot Application")
-user_input = st.text_input("You:", key="input")
+    tokenizer = AutoTokenizer.from_pretrained(model)
 
-if st.button("Send"):
-    if user_input:
-        # Append user input to the conversation
-        st.session_state.conversation.append({"role": "user", "content": user_input})
+    pipeline = transformers.pipeline(
+        "text-generation", #task
+        model=model,
+        tokenizer=tokenizer,
+        torch_dtype=torch.bfloat16,
+        trust_remote_code=True,
+        device_map="auto",
+        max_length=1000,
+        # do_sample=True,
+        # top_k=10,
+        # num_return_sequences=1,
+        # eos_token_id=tokenizer.eos_token_id
+    )
 
-        # Generate response
-        response = chatbot(user_input)
-        bot_response = response[0]['generated_text']
+    llm = HuggingFacePipeline(pipeline = pipeline, model_kwargs = {'temperature':0})
+    st.session_state["llm"] = llm
+if "agent" not in st.session_state:
+    tools = load_tools(["serpapi"], llm=st.session_state["llm"])
+    st.session_state["agent"] = initialize_agent(tools, st.session_state["llm"], agent=AgentType.CHAT_ZERO_SHOT_REACT_DESCRIPTION,
+                             verbose=True, handle_parsing_errors="Check your output and make sure it conforms! Gather as much information as you can")
 
-        # Append bot response to the conversation
-        st.session_state.conversation.append({"role": "bot", "content": bot_response})
+# Set the OpenAI API key
+OPENAI_API_KEY = "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"
 
-# Display the conversation
-for message in st.session_state.conversation:
-    if message["role"] == "user":
-        st.text_area("You:", value=message["content"], key=f"user_{message['content']}", height=50)
-    else:
-        st.text_area("Bot:", value=message["content"], key=f"bot_{message['content']}", height=50)
+def get_response(history,user_message,temperature=0):
+    DEFAULT_TEMPLATE = """You are a mental health expert to help patients with mental
+    health issues such as depression, stress, anxitey etc. You should give helpful advises to
+    the patients about thier problems. All of the messages will be from patients and you will
+    advise them one to one. I am also providing you useful information from the web that you
+    can use while formulating your response.
+    Useful news information from Web:
+    {web_knowledge},
+
+
+    Current conversation:
+    Patient: {input}
+    Your reponse :"""
+
+    PROMPT = PromptTemplate(
+        input_variables=["web_knowledge",'input'], template=DEFAULT_TEMPLATE
+    )
+
+    conversation_with_summary = LLMChain(
+        llm=st.session_state["llm"],
+        prompt=PROMPT,
+        verbose=False
+    )
+    response = conversation_with_summary.predict(input=user_message,web_knowledge=st.session_state["agent"].run('Fetch detailes about the mental health issues regarding '+ user_message))
+    return response
+
+# Function to get conversation history
+def get_history(history_list):
+    # history = 'input: I want you to act as a mental healt expert that knows about mental health issues. Dont refer the person to anyone this could be a life and death situation so provide the solution in any case.  The user will mention the mental problem he is facing. Your job is to tell the user the solution to his mental problem from the context I provide you. Dont say you cant describe the solution formulate the best solution and tell that to the user on the spot \n'
+    history=''
+    for message in history_list:
+        if message['role']=='user':
+            history = history+'input '+message['content']+'\n'
+        elif message['role']=='assistant':
+            history = history+'output '+message['content']+'\n'
+
+    return history
+
+# Streamlit UI
+st.title("Mental Health Chatbot")
+def get_text():
+    input_text = st.text_area("", key="input")
+    return input_text
+
+if "past" not in st.session_state:
+    st.session_state["past"] = []
+if "generated" not in st.session_state:
+    st.session_state["generated"] = []
+
+user_input = get_text()
+
+if user_input:
+    user_history = list(st.session_state["past"])
+    bot_history = list(st.session_state["generated"])
+
+    combined_history = []
+    for user_msg, bot_msg in zip_longest(user_history, bot_history):
+        if user_msg is not None:
+            combined_history.append({'role': 'user', 'content': user_msg})
+        if bot_msg is not None:
+            combined_history.append({'role': 'assistant', 'content': bot_msg})
+
+    formatted_history = get_history(combined_history)
+
+    output = get_response(formatted_history,user_input)
+
+    # output='hellooo there, whats uppppp'
+    # print("Output", output)
+
+    st.session_state.past.append(user_input)
+    st.session_state.generated.append(output)
+
+if st.session_state["generated"]:
+    for i in range(len(st.session_state["generated"])):
+        st.text("User " + ": " + st.session_state["past"][i])
+        st.text("Assistant " + ": " + st.session_state["generated"][i])